All Classes and Interfaces
Class
Description
Shared
ProviderErrorMapper for Google Cloud LLM endpoints — the
Generative Language API (Gemini) and Vertex AI both return the canonical
Google API-Gateway error envelope, shared across every Google Cloud service:Abstract base class for LLM client implementations.
Base class for providers that speak the OpenAI
/chat/completions
wire format verbatim — same request body, same choices[].message
response, same choices[].delta streaming, Bearer auth.Shared
ProviderErrorMapper for providers that clone OpenAI's
chat-completions API at the error-envelope level.Anthropic Claude LLM client implementation.
Builder for creating AnthropicClient instances.
ProviderErrorMapper for Anthropic's Messages API.Translates the OpenAI-flavoured intermediate tool schema into
Anthropic's Messages-API tool shape.
Azure OpenAI Service client implementation.
Authentication type for Azure OpenAI.
Exception thrown when Azure OpenAI content filter rejects the request.
ProviderErrorMapper for Azure OpenAI Service — ninth pilot
mapper.AWS Bedrock LLM client implementation.
ProviderErrorMapper for AWS Bedrock — sixth pilot mapper.Exception thrown when a budget limit is exceeded.
Manages budget limits and alerts for LLM usage.
Wrapper for content that can be cached.
A content block with optional cache control.
Cache control configuration for prompt caching.
Cache control type.
LLM client wrapper that adds semantic caching capabilities.
Builder for CachedLLMClient.
Represents a cached LLM response entry.
Cache usage statistics from an LLM response.
Routes LLM requests based on required capabilities.
Builder for CapabilityRouter.
Decorator that wraps any
LLMClient and emits one
LLMCallLog per call to a configured LLMCallPublisher
(issue #79).Cerebras LLM client.
ProviderErrorMapper for Cerebras — an OpenAI-compatible provider.LLMClient decorator that implements the circuit breaker pattern.
Snapshot of circuit breaker metrics.
Circuit breaker states.
Thrown when a request is rejected because the circuit breaker is open.
Cohere LLM client implementation.
ProviderErrorMapper for Cohere — tenth pilot mapper.Translates the OpenAI-flavoured intermediate tool schema into
Cohere's chat-API tool shape.
LLMClient wrapper that tracks usage costs and enforces budgets.
Routes LLM requests to the cheapest viable provider.
Builder for CostBasedRouter.
Cost statistics.
USD cost breakdown for one LLM call (issue #79).
Interface for tracking LLM API usage costs.
Aggregated cost statistics.
Databricks Mosaic AI Model Serving LLM client.
ProviderErrorMapper for Databricks — an OpenAI-compatible provider.DeepInfra LLM client.
ProviderErrorMapper for DeepInfra — an OpenAI-compatible provider.DeepSeek LLM client.
ProviderErrorMapper for DeepSeek — an OpenAI-compatible provider.Provides embeddings for text to enable semantic similarity search.
The task an embedding is generated for.
Failure detail for a non-successful LLM call (issue #79).
Sealed predicate over an
LLMException describing when a
LLMFallbackRule should fire (issue #89).Always matches — the catch-all "last resort" condition.
Logical AND/OR over child conditions.
Match when the failure's classified
LLMException.ErrorType
is in the configured set.Match when the upstream HTTP status (from
ProviderDetails)
is in the configured set.Match when the failure originated from the named provider AND its
native error code (from the
ProviderDetails envelope) is in
the configured set.LLM router with automatic fallback on failure.
Builder for FallbackRouter.
Fireworks.ai LLM client.
ProviderErrorMapper for Fireworks.ai — an OpenAI-compatible provider.Google Gemini LLM client with multimodal support (text, image, audio, video).
ProviderErrorMapper for Google's Generative Language API (Gemini).Groq LLM client implementation.
ProviderErrorMapper for Groq — seventh pilot mapper.Simple hash-based embedding provider for testing and development.
HuggingFace Inference API client using OpenAI-compatible endpoint.
ProviderErrorMapper for HuggingFace Inference API —
eleventh pilot mapper.In-memory implementation of CostTracker.
In-memory
LLMPricingRegistry populated programmatically via
InMemoryLLMPricingRegistry.Builder (issue #79).In-memory implementation of SemanticCache.
Builder for InMemorySemanticCache.
Extended LLM client interface that supports native JSON mode.
LLMPricingRegistry that loads rates from a classpath JSON file
(issue #79).LLM client wrapper that enforces JSON mode output.
Builder for JsonModeClient.
Exception thrown when JSON parsing fails.
Routes LLM requests to the fastest available provider.
Builder for LatencyBasedRouter.
Latency statistics.
llama.cpp server LLM client.
ProviderErrorMapper for llama.cpp — an OpenAI-compatible provider.LLM-based desire conflict resolver.
Builder for LLMBasedConflictResolver.
One observed LLM call, end-to-end (issue #79).
Mutable builder — collects fields as the call executes
(start → response → finish), then produces the immutable record.
SPI for publishing
LLMCallLog events emitted by
CapturingLLMClient (issue #79).Thrown by an
LLMFallbackChain when every eligible target has
been tried and none succeeded (issue #89).One try against one chain target.
Factory for creating LLM clients from configuration.
ServiceLoader-based implementation of
LLMClientProvider.Phase 1 transport / sampling extras from
LLMSpec (TAN-3637).An
LLMClient that wraps a primary plus a list of
LLMFallbackRules, transparently routing failed calls to the
first matching, not-cooling-down fallback target (issue #89).Fluent configurator.
One entry in an
LLMFallbackChain: a target LLMClient
to try, the FallbackCondition that gates it, and an optional
cooldown window (issue #89).Collects and reports metrics for LLM operations.
Per-provider metrics.
Comprehensive metrics report.
Observer interface for LLM request/response monitoring.
SPI for converting
TokenUsage into a USD CostEstimate
(issue #79).Captured input side of an LLM call (issue #79).
Cache-control metadata for prompt-caching providers (Anthropic,
Google).
Captured output side of an LLM call (issue #79).
One tool-call request from the LLM.
Routes LLM requests to appropriate providers based on configurable strategies.
Records for routing statistics.
LM Studio LLM client.
ProviderErrorMapper for LM Studio — an OpenAI-compatible provider.MiniMax LLM client implementation.
ProviderErrorMapper for MiniMax — twelfth pilot mapper.Mistral AI LLM client implementation.
ProviderErrorMapper for Mistral's chat completions API.Model pricing information for cost calculation.
NVIDIA NIM (NVIDIA Inference Microservices) LLM client.
ProviderErrorMapper for NVIDIA NIM — an OpenAI-compatible provider.Wrapper that adds observability to any LLMClient.
Composes multiple observers into one.
Ollama LLM client implementation for local and cloud deployments.
EmbeddingProvider backed by an Ollama server's /api/embed
endpoint.ProviderErrorMapper for Ollama, the local-first LLM
runtime that runs models on the host machine
(http://localhost:11434 by default).OpenAI LLM client implementation.
ProviderErrorMapper for OpenAI's chat / completions / embeddings
APIs.OpenRouter LLM client implementation.
ProviderErrorMapper for OpenRouter — eighth pilot mapper.Perplexity LLM client.
ProviderErrorMapper for Perplexity — an OpenAI-compatible provider.LLMClient wrapper that adds prompt caching support.
Logs LLM prompts and responses with PII filtering support.
Builder for PromptLogger.
Provider-specific translation from a raw HTTP error response to a typed
LLMException with attached ProviderDetails.Static lookup facade for
ProviderErrorMappers registered via
ServiceLoader.Alibaba Qwen Cloud (DashScope) LLM client.
ProviderErrorMapper for DashScope — an OpenAI-compatible provider.Replicate LLM client.
ProviderErrorMapper for Replicate.Represents the desired output format for LLM responses.
Response format type.
Routes LLM requests using round-robin load distribution.
Builder for RoundRobinRouter.
Generates JSON Schema from Java classes using reflection.
Semantic cache for LLM responses based on prompt similarity.
Cache statistics.
Result of a similarity search.
Default
LLMCallPublisher that emits one structured SLF4J line
per LLM call (issue #79).Streaming-specific timing metrics for one LLM call (issue #79).
Routes LLM requests to the most appropriate model based on task type.
Builder for TaskBasedRouter.
Task routing statistics.
Classifies user prompts into task types for intelligent model routing.
Builder for TaskClassifier.
Classification result with confidence score.
Defines task types for intelligent model routing.
Tencent Hunyuan LLM client.
ProviderErrorMapper for Hunyuan — an OpenAI-compatible provider.Together.ai LLM client.
ProviderErrorMapper for Together.ai — an OpenAI-compatible provider.Token-count breakdown for one LLM call (issue #79).
Snapshot of the tool set advertised to the LLM on one call (issue #79).
Single canonical entry point for building a
ToolSurface from
the raw List<Map<String,Object>> tool shape every framework
caller passes around (issue #79).Record of a single LLM API call usage.
Google Cloud Vertex AI LLM client (enterprise Gemini).
ProviderErrorMapper for Google Cloud Vertex AI.vLLM LLM client.
ProviderErrorMapper for vLLM — an OpenAI-compatible provider.IBM watsonx.ai LLM client.
ProviderErrorMapper for IBM watsonx.ai.OpenAI Whisper client for audio transcription and translation.
Response format for transcription/translation.
Transcription options.
Transcription result with optional metadata.
Translation options.
xAI Grok LLM client.
ProviderErrorMapper for xAI Grok — an OpenAI-compatible provider.01.AI Yi LLM client.
ProviderErrorMapper for Yi — an OpenAI-compatible provider.ZhipuAI (Z.AI/Zhipu) LLM client implementation.
ProviderErrorMapper for ZhipuAI (Z.AI / GLM) — thirteenth
and final pilot mapper for the original SPI matrix.